Intelligent detection method and device for turnout spatial structure parameters

Abstract

This invention discloses an intelligent detection device and method for turnout spatial structure parameters. Utilizing laser scanning and multi-sensor fusion technology, and integrating an electromechanical structure, it achieves effective detection of turnout spatial structure parameters, meeting the rapid and comprehensive detection needs of common single turnouts. It solves the drawbacks of low efficiency and numerous missed detections in manual inspection, significantly improving detection efficiency and providing automated and intelligent detection equipment and technical means for railway engineering sites.

Description

Technical Field

[0001] This invention relates to a detection method and apparatus, and more particularly to an intelligent detection method and apparatus for turnout spatial structure parameters. Background Technology

[0002] As a key piece of equipment on railway tracks, railway turnouts play a vital role in train steering and track connection. The accuracy of their spatial structural parameters directly affects the safety and smoothness of train operation. With the continuous expansion of my country's railway network and the sustained increase in train speeds, especially the rapid development of high-speed railways, higher demands are placed on turnout condition inspection. Traditional turnout inspection mainly relies on manual measurement using simple tools such as track gauges and levels. This method is not only inefficient and has limited accuracy, but also fails to fully reflect the three-dimensional spatial condition of the turnout, and can no longer meet the needs of modern railway operation and maintenance.

[0003] In recent years, with the rapid development of sensor technology, machine vision, laser measurement, and artificial intelligence, track geometry detection technology has made significant progress. Advanced measurement equipment such as total stations, inertial measurement units (IMUs), and laser scanners have begun to be applied in track inspection, providing new technical means for measuring turnout spatial parameters. Meanwhile, computer vision technology, through image processing and analysis, can acquire track geometric features non-contactly; machine learning algorithms can extract useful information from massive amounts of detection data, enabling intelligent assessment of turnout conditions. However, existing technologies are mostly designed for ordinary tracks, and research on intelligent detection methods specifically for the special structure of turnouts remains relatively insufficient. The track structure in the turnout area is complex, consisting of switches, connecting parts, frogs, and guard rails, with strict geometric relationships between these components. Its spatial parameters include multiple dimensions such as gauge, level, elevation, direction, offset, inspection interval, and guard rail distance, and these parameters vary depending on the turnout's direction. Furthermore, the turnout area contains various track surface discontinuities and geometric abrupt changes, such as switch reduction values ​​and guard rail buffer sections, making traditional continuous track inspection methods difficult to apply directly. These structural features present unique challenges to the accurate measurement of turnout spatial parameters, necessitating the development of specialized detection algorithms and technical solutions.

[0004] Modern railway operations have placed new demands on turnout inspection, requiring "high precision, high efficiency, and intelligence": inspection accuracy must reach sub-millimeter levels to meet the needs of high-speed operation; the inspection process should minimize disruption to normal traffic, enabling rapid and even dynamic measurements; and the inspection system must possess automation and intelligence capabilities, processing data in real time and outputting evaluation results. These demands have spurred the development of an "intelligent detection method for turnout spatial structure parameters." By integrating multi-source sensor data, establishing dedicated algorithm models, and introducing artificial intelligence technology, this method achieves comprehensive, accurate, and efficient detection of turnout spatial parameters, providing scientific condition assessment and maintenance decision support for railway engineering departments. Multi-sensor fusion technology will improve the robustness and accuracy of the measurement system; three-dimensional laser scanning and photogrammetry technologies can acquire complete spatial information of the turnout; deep learning-based image recognition algorithms can automatically extract key turnout features; and digital twin technology can achieve virtual simulation of turnout status and predictive maintenance. Against this backdrop, the invention patent for the "intelligent detection method for turnout spatial structure parameters" represents the forefront of technology in this field. Its widespread application will significantly improve the intelligence level of railway turnout inspection in my country, ensuring the safe and efficient operation of railway transportation.

[0005] As a key piece of equipment in railway tracks, the accuracy of the spatial structural parameters of turnouts directly affects the safety and stability of train operation. Currently, manual inspection mainly employs a combination of traditional mechanical tools and digital instruments, covering the measurement of core parameters such as track gauge, level, elevation, direction, and offset. The following are the main technical methods and equipment:

[0006] 1. Traditional mechanical measuring tools

[0007] (1) Track gauge: The most basic inspection tool used to measure track gauge and level. It obtains the distance between the two rails through a rail clamping structure and uses a spirit level to determine the lateral height difference of the rail surface. The accuracy of ordinary track gauges is ±1mm, and high-precision models can reach ±0.3mm.

[0008] (2) Support gauge: It is used to measure the support of the guide curve. The reading is obtained by using a fixed reference point and a movable measuring rod with a scale. The accuracy is usually ±0.5mm.

[0009] (3) String line: Use a 10m / 20m steel string line with a feeler gauge to detect the track direction and height deviation by measuring the distance between the string line and the rail web.

[0010] (4) Feeler gauge: used for precise measurement of millimeter-level gaps such as the tightness between the switch rail and the main rail, and the gap between the top rail and the main rail.

[0011] 2. Digital measurement tools

[0012] (5) Total station: It uses the polar coordinate method to acquire three-dimensional coordinates and can accurately measure the spatial position of key points of turnouts. The planar accuracy can reach ±1mm and the elevation accuracy is ±0.5mm. It needs to be used with a special target and measurement scheme.

[0013] (6) Laser collimator: It uses the linear characteristics of laser to detect track deviation and can display the track straightness offset in real time. It is suitable for long-distance continuous measurement.

[0014] (7) Electronic track gauge: integrates displacement sensor and inclinometer, automatically calculates track gauge, level and triangular pit parameters, data storage function facilitates trend analysis, and measurement accuracy reaches ±0.2mm.

[0015] (8) Digital offset measuring instrument: Laser ranging technology is used to replace the traditional mechanical ruler, improving the efficiency and accuracy of offset detection.

[0016] (9) Degradation measuring instrument: The displacement sensor is used to replace the traditional vernier caliper to improve the detection accuracy of the degradation of the center rail and switch rail.

[0017] The aforementioned traditional turnout inspection techniques have long relied on manual operation of simple tools such as track gauges and levels. While these techniques offer advantages like low cost and ease of operation, their inefficiency (approximately 80 minutes per inspection) and limited accuracy (gauge measurement error of ±1mm) are becoming increasingly apparent. Manual inspection is limited by the operator's experience, resulting in highly subjective and poor repeatability of measurement results. In particular, the identification rate for hidden defects such as cracks and loose bolts is less than 30%, failing to meet the stringent accuracy requirements of modern railways. Traditional tools can only obtain local parameters at discrete points, failing to comprehensively reflect the three-dimensional spatial state and dynamic deformation characteristics of the turnout. For example, deviations of 0.3mm in switch rail fit are often missed. Furthermore, the inspection process frequently occupies "track windows," leading to a 15%-20% decrease in track capacity. Paper records also create data silos, delaying defect tracing by over 48 hours, severely hindering long-term tracking and trend prediction of turnout conditions. These technical bottlenecks are particularly prominent in high-speed rail scenarios. The demand for 0.2mm-level accuracy and real-time monitoring in operating environments above 200km / h exposes the systemic shortcomings of traditional methods.

[0018] The limitations of traditional inspection methods are further exacerbated by the technical blind spots and safety hazards inherent in manual inspections. Subjective judgment methods such as visual inspection and tapping / listening lack standardized criteria, and the skill differences among operators lead to a fault misjudgment rate as high as 15%, with inspection accuracy deteriorating drastically at night or in inclement weather. The lack of dynamic parameter capture capabilities prevents the real-time acquisition of critical data such as track gauge changes, and track maintenance requires 3-4 people working together, resulting in low efficiency (inspection speed <10m / h) and causing 72% of inspection-related injuries and fatalities to occur in the turnout area. Outdated data management makes it difficult to model and analyze long-term evolution patterns such as annual rail wear of 0.2-0.5mm, leading to severe delays in fault warning and response. In contrast, while digital equipment can achieve sub-millimeter level inspection accuracy, it suffers from limitations such as limited functionality and high professional training requirements. Summary of the Invention

[0019] To address the shortcomings of existing technologies, this invention discloses an intelligent detection device for turnout spatial structure parameters, the technical solution of which is as follows:

[0020] A smart detection device for turnout spatial structure parameters, characterized in that it comprises:

[0021] Running frame: T-shaped layout with rigid connection between crossbeams and longitudinal beams, installed at the bottom.

[0022] Running wheels, tensioning wheel assembly, and guide wheel assembly: clamp the rail head to achieve directional movement; the running wheels are connected to a high-precision encoder to record mileage.

[0023] Track gauge compensation wheel and measuring wheel: These are attached to the inner side of the rail 16mm below the top of the rail, and are used to measure the track gauge during travel.

[0024] Horizontal compensation wheel and track gauge compensation wheel: compensate for track unevenness through a spring mechanism;

[0025] Laser scanning module:

[0026] The crossbeam laser scanning module and the longitudinal beam laser scanning module are fixed to the traveling frame via the laser scanning module backplate;

[0027] The horizontal and vertical beam laser scanning module contains three laser cameras (blue-red-blue) with wavelengths of 405nm±5nm, 650nm±5nm, and 405nm±5nm, respectively.

[0028] Data acquisition module: integrates a MASTER 810 controller and a data acquisition card. The MASTER 810 controller controls the laser camera group in parallel via a network cable; the data acquisition card connects to displacement gauges, tilt sensors, fiber optic gyroscopes, and temperature sensors via data transmission cables.

[0029] The photoelectric encoder's mileage pulse signal is connected to the MASTER 810 controller and data acquisition card via data transmission cables, triggering the laser camera to scan, as well as the displacement meter, tilt sensor, fiber optic gyroscope, and temperature sensor.

[0030] Data analysis and processing unit: built-in computing unit, PCL threshold segmentation algorithm and Kalman filter program;

[0031] Positioning structure: The back plate is locked to the traveling frame by transverse positioning pins, longitudinal positioning pins and positioning discs, with a repeatability error of ≤0.1mm.

[0032] This invention also discloses an intelligent detection method for turnout spatial structure parameters, characterized by the following steps: two parameters are collected synchronously and in parallel:

[0033] Track turnout structural parameter acquisition

[0034] Step S1: Synchronous acquisition of multi-source data:

[0035] The traveling frame moves along the track, and the encoder outputs a trigger signal every 0.125m;

[0036] A laser camera synchronously scans and generates a 3D point cloud of track components in the turnout area (resolution 0.05mm), with a cross-sectional interval of 5mm.

[0037] Fiber optic gyroscopes collect pitch and yaw angles, tilt sensors collect roll angles, and displacement gauges collect the lateral displacement distance of the measuring wheel.

[0038] Step S2: Point cloud preprocessing:

[0039] Statistical outlier removal (SOR): Remove noise points with a distance > μ+3σ from a neighborhood radius of 0.1dm.

[0040] Motion compensation: Correcting point cloud pose by combining encoder odometer data and gyroscope data;

[0041] Step S3: Feature section extraction:

[0042] PCL threshold segmentation algorithm: Set a threshold k based on the horizontal coordinate interval of the point cloud data. When k is greater than 0.3mm, it is used as the profile segmentation point.

[0043] ICP template matching: Based on the turnout design CAD model, iteratively optimize point cloud registration (convergence condition: displacement error < 0.1mm).

[0044] Step S4: Structural parameter calculation: including reduction value calculation and flange groove width;

[0045] Step S5: Track geometry parameter processing:

[0046] The FIR low-pass filter RLS algorithm is used to process ultra-high data and output ultra-high values.

[0047] Beneficial effects

[0048] 1. Improve the efficiency and automation level of turnout inspection.

[0049] By using multi-sensor fusion and automated scanning technology, including laser cameras, inertial measurement units, tilt measurement units, and displacement gauges, turnout parameters can be collected quickly and automatically, reducing manual intervention.

[0050] 2. Improve measurement accuracy and reliability

[0051] Employing a sub-millimeter-level laser ranging high-precision sensor, ICP point cloud registration, and PCL threshold segmentation algorithm, environmental interference is eliminated to ensure high-precision measurement of key parameters.

[0052] 3. Implement over-limit early warning

[0053] By combining real-time data transmission, the dynamic deformation data of turnouts is detected online, and the risk of faults such as switch rail jamming and abnormal track gauge is predicted through threshold early warning.

[0054] Before measurement, the turnout gauge design parameters in the database are called in advance, and the gauge value is compared in real time according to the mileage. When the measured value is greater than or less than the design value by 3mm, an alarm is triggered.

[0055] 4. Reduce operation and maintenance costs and security risks

[0056] By replacing high-risk manual operations (such as nighttime track inspections) with intelligent systems, the efficiency of operations during track maintenance windows can be improved, ensuring the healthy service of turnouts. Attached Figure Description

[0057] Figure 1 Schematic diagram of intelligent detection device for turnout spatial structure parameters;

[0058] Figure 2 This is a schematic diagram of the scanning field of view of the laser scanning module; Figure 3 This is a schematic diagram of the backplane of the laser scanning module;

[0059] Figure 4 This is the main view of the walkway frame;

[0060] Figure 5 Left view of the walkway frame;

[0061] Figure 6 This is the main view of the walkway frame;

[0062] Figure 7 This is the main view of the splicing surface of the horizontal and vertical beams;

[0063] Figure 8Flowchart of the track geometry measurement system operation;

[0064] Figure 9 Flowchart of the turnout structural parameter measurement system;

[0065] Figure 10. Flowchart of orbit geometry solution;

[0066] Figure 11 Flowchart of algorithm for turnout structural parameters;

[0067] Figure 12 Comparison of rail profile curves;

[0068] Figure 13 Comparison of rail profile phase coordinates;

[0069] Figure 14 Euclidean distance deviation results;

[0070] Figure 15 A comparative diagram of track geometry (gauge) measurement results, where (a) gauge measurement value, and (b) track deviation;

[0071] Figure 16 A comparative diagram of track geometry (horizontal) measurement results, where (a) horizontal measurement value; (b) horizontal deviation;

[0072] Figure 17 A comparative diagram of track geometry (orbital alignment) measurement results, where (a) orbital alignment measurement value; (b) orbital alignment deviation;

[0073] Figure 18 This is a schematic diagram comparing the results of track geometry (elevation) measurements, where (a) elevation measurements and (b) elevation deviations.

[0074] The components include: 1. Crossbeam laser scanning module; 2. Traveling frame; 3. Data analysis and processing unit; 4. Push rod; 5. Data acquisition module; 6. Longitudinal beam laser scanning module; 7. Laser camera 1 (blue); 8. Laser camera 2 (red); 9. Laser camera 3 (blue); 10. Envelope of track component measurement range in turnout area; 11. Laser scanning module backplate; 12. Lateral positioning pin; 13. Positioning disk; 14. Longitudinal positioning pin; 15. Tensioning wheel assembly; 16. Traveling wheel; 17. Guide wheel assembly; 18. Longitudinal beam; 19. Horizontal compensation wheel; 20. Track gauge compensation wheel; 21. Inclination sensor; 22. Measuring wheel; 23. Crossbeam traveling wheel; 24. Displacement gauge; 25. Crossbeam; 26. Fiber optic gyroscope; 27. Battery; 28. High-precision encoder; 29. ​​MAETER. 810 and micro switch 30, data analysis and processing machine 31, horizontal and vertical beam locking bolts 32, data acquisition card 33, vertical locking positioning pin of horizontal and vertical beams 34, horizontal and vertical beam positioning plate 35, horizontal and vertical beam fastening threaded hole 36, horizontal and vertical beam horizontal locking positioning pin. Detailed Implementation

[0075] Example 1

[0076] The intelligent detection device for turnout spatial structure parameters mainly consists of a crossbeam laser scanning module, a longitudinal beam laser scanning module, a traveling frame, a data analysis and processing unit, a data acquisition module, and push rods, as shown in the figure.

[0077] 1) Horizontal and longitudinal beam laser scanning module

[0078] The horizontal and vertical beam laser scanning modules are mainly used for scanning the profile of track components in the turnout area, generating point cloud data of the track component profile, and calculating track structure parameters such as downsizing, wheel flange groove width, and rail wear. Each laser scanning module consists of 3 laser cameras, using a blue-red-blue light combination.

[0079] like Figure 2 As shown, the laser camera's field of view layout is such that at the rail head, the blue light camera's field of view does not overlap, avoiding interference from reflections of the same color light and ensuring the stability and reliability of the measurement data.

[0080] Hardware performance is not correlated with the effect: the "repeated disassembly and assembly accuracy of 0.1mm" of the travel frame has not been verified in terms of how it ensures measurement accuracy.

[0081]

[0082] The key parameters of the algorithm have not been validated: the engineering basis for the forgetting factor λ=0.98 and the RLS initialization parameter δ=0.02 in the adaptive filtering is not provided.

[0083] The laser scanning module backplate is bolted to the running frame. Three-point positioning ensures repeatable assembly and disassembly accuracy of 0.1mm, enabling rapid assembly and disassembly. The specific operation steps are as follows:

[0084] Step 1: Assemble the traveling frame 2. The crossbeams 25 and longitudinal beams 18 are locked with crossbeam and longitudinal beam locking bolts 31. The bolts are screwed into the fastening threaded holes of the crossbeams and longitudinal beams to secure them. The vertical locking positioning pins 33 of the crossbeams and longitudinal beams ensure vertical positioning of the crossbeams and longitudinal beams. The horizontal locking positioning pins 36 of the crossbeams and longitudinal beams ensure horizontal positioning of the crossbeams and longitudinal beams. The positioning discs 34 of the crossbeams and longitudinal beams ensure circumferential locking positioning.

[0085] Step 2: Install the laser scanning modules for the horizontal and vertical beams. The back plate 11 of the laser scanning module is connected to the positioning discs 13 on the horizontal and vertical beams, respectively. Bolts are screwed into the threaded holes in the center of the positioning discs to secure the connection. The horizontal positioning pin 12 and the vertical positioning pin 14 ensure the horizontal and vertical positioning accuracy of the back plate of the laser scanning module.

[0086] Step 3: Install the push rod and set up the data analysis and processing machine.

[0087] 2) Traveling frame

[0088] The running frame is mainly composed of running wheels, guide wheels, measuring wheels, compensation wheels, displacement gauges, tilt sensors, fiber optic gyroscopes, batteries, etc.

[0089] The main function of the running frame is:

[0090] (1) The running frame moves on the track in the turnout area as the carrier of the laser measurement module. The mileage is counted by the built-in encoder in the running frame, and the trigger signal is provided to the laser measurement module and the data acquisition module at the same time.

[0091] (2) The running frame is used for the precise measurement of track geometry parameters such as gauge, level, direction, and elevation in the turnout area.

[0092] Table: Measurement Error of Track Geometric Parameters by Intelligent Detection Device for Turnout Spatial Structure Parameters

[0093]

[0094] 3) Data analysis and processing machine

[0095] The data analysis and processing unit is mainly used for extracting and analyzing the feature points of the cross-sectional profile of the turnout area, calculating the turnout structural parameters, and displaying the track geometry data and real-time rail profile of the turnout area.

[0096] 4) Push rod

[0097] Used to propel the traveling frame to move on the rails.

[0098] 5) Data Acquisition Module

[0099] It mainly consists of a data acquisition board, a MASTER 810, and a micro switch. It is used for A / D conversion and uploading of data from displacement gauges, temperature sensors, encoders, tilt sensors, fiber optic gyroscopes, etc., as well as for synchronous triggering of laser measurement modules and uploading of point cloud data.

[0100] The point cloud data processing algorithm is as follows:

[0101] (1) SOR noise reduction: with point p i For example, in the processing of:

[0102] 1) Neighborhood definition: r is the radius of the neighborhood, based on the empirical value r = 30mm; |N i | represents the total number of points in the neighborhood (including p) i ),

[0103]

[0104] 2) Calculate the distance from the midpoint of the set to p. i Distance set |Di |:

[0105]

[0106] 3) Calculate the mean distance and standard deviation :

[0107]

[0108]

[0109] 4) Set the outlier threshold: k is the standard deviation multiplier, k=5.

[0110]

[0111] 5) Calculate |N i | Center of mass:

[0112]

[0113] 6) Calculate the distance from the point to the centroid:

[0114]

[0115] 7) Determine the filtering conditions: If so, then delete this point.

[0116] (2) Profile segmentation: (PCL threshold segmentation algorithm)

[0117] a) Track vertex :

[0118] Take the filtered cross-sectional profile point cloud data set from the beam laser scanning module |M i |, find |M i The point with the largest y-value in | ( );

[0119] b) Track gauge point :

[0120] like The vertical axis in ,but = There are two gauge points for a single cross-section profile, which are respectively determined by the magnitude of their x-coordinates. ,

[0121] c) Calculate the data set |Q 16mm below the orbital vertex i |, |Q i | |M i|,

[0122] If point ( , ), ,but |Q i |;

[0123] d) The data set 16mm below the track vertex |Q i | Slicing process: Subtract the x-coordinates of two adjacent points; if the difference is greater than 0.3, then the point is used as the boundary of the set; ( , ), ( , ... ( , ), but |O 1 i |, |O 2 i |

[0124] (3) Target recognition:

[0125] 1) Basic track identification: 50 consecutive cross-sectional profiles | Q i |Q i+1 |……|Q i+50 | in ( , ), ( , ), ,like , then |Q i |Q i+1 |……|Q i+50 | It is determined to be a basic orbital point cloud dataset.

[0126] 2) Switch rail recognition: 50 consecutive cross-sectional profiles | Q i |Q i+1 |……|Q i+50 | in ( , ), ( , ),like And since there are two data slice sets, then |O i 2∣∣O i+1 2 |…… |Oi+50 2 | If it is determined to be a switch rail and its working state is close contact, then the side rail of the longitudinal beam laser scanning module is also a switch rail and its working state is repulsion.

[0127] 3) Guardrail identification: Inside the basic rail, with a width greater than 40, and a distance between 60 and 90 from the basic rail; the cross-sectional profile slice dataset consists of 2 samples. For |O 1 i The point with the smallest x-coordinate in | , They are respectively |O 2 i The points with the largest and smallest x-coordinates in |, if and , then |O i 2 | is the point cloud dataset for the guardrail.

[0128] 4) Heart Track Recognition: Set 50 continuous profiles | O 1 i ∣∣O 1 i+1 |……|Q 1 i+50 Calculate the slope of the points on the outer side of the |; a slope greater than 0.4 indicates the center track. , ... Sets |O 1 i ∣∣O 1 i+1 |……|Q 1 i+50 For the point with the largest x-coordinate in |, fit the x-coordinate with a slope k. If k > 0.4, then |O 1 i ∣∣O 1 i+1 |……|Q 1 i+50 | is the point cloud dataset for the orbital trajectory.

[0129] 5) Wing rail identification: After identifying the wing rail, the corresponding |O 2 i ∣∣O 2 i+1 |……|Q 2 i+50 | That is, the wing-rail point cloud dataset.

[0130] (4) Feature point extraction:

[0131] The point with the highest y-coordinate value in each rail component dataset is used as reserve data for calculating the reduction value. Points in each rail component dataset with the same y-coordinate as the basic rail gauge point are used as reserve data for calculating the flange groove width.

[0132] (5) Calculation of the reduction value:

[0133] The guard rail elevation value 'e' is the difference between the highest point of the guard rail and the highest point of the main rail.

[0134]

[0135] The drop value f of the airfoil is the difference between the highest point of the airfoil and the highest point of the airfoil.

[0136]

[0137] The switch rail reduction value g is the difference between the highest point of the main rail on the close-fitting side of the switch rail and the highest point of the switch rail.

[0138]

[0139] (6) Calculation of rim groove width:

[0140] guard rail flange groove width h

[0141]

[0142] switch rail flange groove width k

[0143]

[0144] Heart rail rim groove width q

[0145] If the highest point of the center rail is higher than the highest point of the wing rail, then the width q of the center rail flange groove in this cross-sectional profile is the difference between the abscissa of the center rail gauge point and the abscissa of the wing rail.

[0146]

[0147] If the highest point of the wing rail is higher than the highest point of the center rail, then the width q of the center rail flange groove in this cross-sectional profile is the difference between the abscissa of the wing rail gauge point and the abscissa of the center rail.

[0148]

[0149] 6) Data extraction and processing algorithm

[0150] (1) PCL threshold segmentation algorithm: extracting feature sections

[0151] The algorithm flow for extracting key structural parameters of turnout feature sections is as follows:

[0152] 1) Data preprocessing

[0153] • Noise reduction filtering:

[0154] Statistical Outlier Removal (SOR): For each point, calculate the mean distance μ and standard deviation σ of its neighborhood (radius 0.1 dm), and remove outliers whose distance is greater than μ + 3σ.

[0155] 2) Feature point extraction

[0156] For point cloud data of different structural components in the turnout area, RANSAC was used to fit a linear model (maximum iterations 20 times, interior point threshold 0.3 mm) to extract the highest point of the rail top and the gauge point.

[0157] 4) Calculation of turnout structural parameters

[0158] • Calculation of reduction value

[0159] Take the turnout cross-section data at the same mileage and calculate the switch rail drop. The calculation method is the difference between the ordinate of the base rail vertex and the switch rail vertex. The calculation formula is as follows:

[0160]

[0161] Similarly, the calculation method for the height difference between the center rail and the guard rail is the same.

[0162] · Calculation of rim groove width

[0163] Using turnout cross-sectional data from the same mileage, calculate the width of the switch rail flange groove. The calculation method is the distance between the base rail gauge point and the horizontal line connecting the switch rail. The calculation formula is as follows:

[0164]

[0165] Similarly, the calculation method for the width of the wheel flange groove of the center rail and guard rail is the same.

[0166] (2) Adaptive filtering algorithm for track geometry parameters

[0167] The track geometry adaptive filtering algorithm is mainly used to process the measurement data of track geometry parameters hyperelevation, in order to remove noise and extract useful information. Its process is as follows:

[0168] 1) Track geometry data acquisition and input

[0169] Raw data, including parameters such as track gauge, elevation, horizontality, and orientation, is collected using tilt sensors, fiber optic gyroscopes, and displacement compensation sensors. Environmental factors that may affect track geometry, such as temperature, are also recorded simultaneously.

[0170] 2) Data preprocessing

[0171] Outlier removal: Use a 15-point sliding window to detect and remove obvious outliers.

[0172] 3) FIR low-pass filter RLS algorithm

[0173] This adaptive filtering algorithm was developed primarily to improve the accuracy and stability of track superelevation measurements in track geometry surveying. It employs a low-pass filtering method to ensure data stability.

[0174] Filter structure: Transverse FIR filter: Low-pass filter.

[0175] Introduction to the FIR low-pass filter RLS algorithm for updating filter coefficients:

[0176] The filter order is 6, the forgetting factor λ = 0.98, and the normalization parameters are... The weight vector is 0.02. It is a 6×1 coefficient matrix, and the inverse correlation matrix is... It is a 6×6 adaptive matrix. The original input data is extremely high. This is the filtered, ultra-high output data.

[0177] initialization: , ,in A 6×6 identity matrix

[0178] Construct a reference vector:

[0179] Calculate the prior error:

[0180] Calculate the gain vector:

[0181]

[0182] Update the inverse correlation matrix:

[0183]

[0184] Update filter weights:

[0185]

[0186] Output filtered ultra-high data:

[0187]

[0188] 7) The accuracy of track structure parameter detection in the turnout area can reach [a certain level].

[0189] Intelligent detection device for turnout spatial structure parameters; accuracy of turnout structure parameter measurement.

[0190]

[0191] Intelligent detection device for turnout spatial structure parameters; accuracy of track geometric parameter measurement.

[0192]

[0193] For the operation procedure of the track geometry measurement system, please refer to Figure 8 As shown: Start measurement → Encoder trigger → Data acquisition (IMU + displacement gauge + tilt sensor + displacement gauge synchronous trigger) → Data acquisition card (track gauge, level, track alignment, elevation calculation) → Micro switch (digital signal) → Data analysis and processing unit (FIR filtering) → Track geometry parameter display.

[0194] For the operation procedure of the turnout structural parameter measurement system, please refer to [link / reference]. Figure 9 As shown: Start measurement → Encoder trigger → MASTER810 (6 cameras trigger simultaneously) → Horizontal and longitudinal beam laser modules scan rail profile → Micro switch → Transmit to data processor → Data analysis processor → Profile segmentation → Feature point extraction → Output reduction value / rim groove width.

[0195] Example 2

[0196] Based on the intelligent detection device for turnout spatial structure parameters shown in Embodiment 1, this invention also discloses an intelligent detection method for turnout spatial structure parameters, comprising the following steps: two parameters are collected synchronously and processed in parallel:

[0197] Track turnout structural parameter acquisition

[0198] Step S1: Synchronous acquisition of multi-source data:

[0199] The traveling frame (2) moves along the track, and the encoder outputs a trigger signal every 0.125m;

[0200] A laser camera synchronously scans and generates a 3D point cloud of track components in the turnout area (resolution 0.05mm), with a cross-sectional interval of 5mm.

[0201] The fiber optic gyroscope (1) collects pitch and yaw angles, the tilt sensor (7) collects roll angles, and the displacement meter (10) collects the lateral displacement distance of the measuring wheel (22).

[0202] Step S2: Point cloud preprocessing:

[0203] Statistical outlier removal (SOR): Remove noise points with a distance > μ+3σ from a neighborhood radius of 0.1dm.

[0204] Motion compensation: Correcting point cloud pose by combining encoder odometer data and gyroscope data;

[0205] Step S3: Feature section extraction:

[0206] PCL threshold segmentation algorithm: Set a threshold k based on the horizontal coordinate interval of the point cloud data. When k is greater than 0.3mm, it is used as the profile segmentation point.

[0207] ICP template matching: Based on the turnout design CAD model, iteratively optimize point cloud registration (convergence condition: displacement error < 0.1mm).

[0208] Step S4: Structural parameter calculation: including reduction value calculation and flange groove width;

[0209] Step S5: Track geometry parameter processing:

[0210] The FIR low-pass filter RLS algorithm is used to process ultra-high data and output ultra-high values.

[0211] Example 3

[0212] Measured data and error analysis on typical turnouts (such as No. 18 turnout with 60kg / m rail).

[0213] A verification test was conducted on a No. 18 turnout with 60 kg / m rail at the National Railway Test Center. The test results of the prototype were compared with those of the Mini Prof rail profiler. Figure 12 Compare different angles of the rail profile, such as... Figure 13 As shown. Calculate the Euclidean distance deviation; the calculation results are as follows. Figure 14 As shown in the figure. The calculation results show that the Euclidean deviation of the profile measured by the prototype and MiniProf is controlled within 0.1 mm, and the rail profile data measured by the prototype is reliable.

[0214] (1) Turnout structural parameters

[0215] The turnout structural parameters measured by the prototype turnout condition inspection instrument were compared with the results of manual measurement, as shown in Tables 1 and 2. The comparison of experimental data with manual measurement results shows that the measurement accuracy of rail component spacing (rim groove width) and rail component height difference (reduction value) reaches sub-millimeter level (0.3mm). The results demonstrate that the prototype measurement method is feasible as an alternative to manual inspection.

[0216] Table 1 Comparison of measurement results of turnout structural parameters (reduction value)

[0217]

[0218] Table 2 Comparison of measurement results of turnout structural parameters (rim groove width)

[0219]

[0220] (2) Track geometric parameters

[0221] The track geometry parameters measured by the prototype turnout condition inspection instrument were compared with the results of manual measurements. Figures 15-18The track alignment and elevation detection chord lengths are selected from the 10 m chord length for evaluating static geometric irregularities of the track as specified in TG / GW 102—2019 "Rules for Repairing Conventional Speed ​​Railway Lines". Comparison of experimental data with manual measurement results shows that the accuracy of the track geometric parameter measurements by the prototype turnout condition checker meets the maximum permissible error requirements for gauge, superelevation, elevation, and track alignment of the Class 0 track checker in Chapter 5.7 of TB / T 3147-2020 Railway Track Checker.

[0222] After practical application at Hefei East Railway Station and Dunhua Station, the prototype turnout condition inspection instrument, operated by one person, took an average of 15-25 minutes to complete the spatial position parameter detection of a set of turnouts, with a data interval of 5mm. Three personnel from the on-site engineering department conducted the spatial position parameter detection of the turnouts at 1-meter data intervals, taking approximately 1-1.5 hours to detect the spatial position parameters of a single No. 18 turnout. The tests demonstrate that the prototype turnout condition inspection instrument provides more comprehensive data and its detection efficiency is more than three times higher than traditional manual measurement.

[0223] This invention is based on a three-dimensional reconstruction technology of the spatial structure of turnout areas using multi-source data fusion (laser scanning + fiber optic gyroscope + encoder). It combines PCL threshold segmentation algorithm to segment feature sections and extract key structural parameters of the turnout. The track geometry parameter detection frame in the turnout area adopts a T-shaped layout and a contact measurement method. It utilizes multi-sensor fusion technology + FIR low-pass filtering RLS algorithm to obtain stable and reliable data results. The device adopts a split structure, supports rapid reassembly within 2 minutes, and integrates a portable detection terminal (including high-precision sensor module and embedded processor) between each split module. The data acquisition results support wireless data transmission and local / cloud collaborative analysis.

[0224] This invention utilizes laser scanning and multi-sensor fusion technology, integrating an electromechanical structure, to achieve effective detection of turnout spatial structural parameters. It meets the rapid and comprehensive detection needs of common single turnouts, solves the drawbacks of low efficiency and numerous missed detections in manual inspection, significantly improves detection efficiency, and provides automated and intelligent detection equipment and technical means for railway engineering sites.

[0225] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for intelligent detection of turnout spatial structure parameters, the method being based on the following intelligent detection device for turnout spatial structure parameters: Running frame: T-shaped layout with rigid connection between crossbeams and longitudinal beams, installed at the bottom. Traveling wheels and guide wheels: clamp the rail head to achieve directional movement; Measuring wheel: fits against the inner wall of the rail and connects to a high-precision encoder to record mileage; Horizontal compensation wheel and track gauge compensation wheel: They compensate for track unevenness through a spring mechanism; Laser scanning module: The crossbeam laser scanning module and the longitudinal beam laser scanning module are fixed to the running frame through the back plate; each module contains three laser cameras: blue-red-blue, with wavelengths of 405nm±5nm, 650nm±5nm, and 405nm±5nm, respectively. Data acquisition module: integrates a MASTER810 controller, connected via an aviation connector, and includes a displacement gauge, tilt sensor, fiber optic gyroscope, and temperature sensor; synchronously receives encoder odometer pulses to trigger laser camera scanning; Data analysis and processing unit: Built-in edge computing unit, running ALOAM algorithm and adaptive filtering program; Positioning structure: The back plate is locked to the traveling frame by transverse positioning pins, longitudinal positioning pins and positioning discs, with a repeatability error of ≤0.1mm; characterized in that The intelligent detection method for turnout spatial structure parameters includes the following: The data acquisition module performs point cloud data processing algorithms, including contour segmentation, and the steps are as follows: a) obtaining track vertices : Obtaining a cross-section profile point cloud data set |M of a certain section of a filtering post-beam laser scanning module i Finding a rail top point with the maximum y value in the cross-section profile point cloud data set |M i , wherein the y value is a vertical coordinate value in the point cloud data​​ b) obtaining gauge points : If the ordinate , then = , the single section profile gauge point has two, according to the size of the abscissa respectively , ; wherein, the ordinate yi is the y value, ytop is the y value of the rail top point Ptop. c) Calculate the 16 mm data set below the top of the rail | Q i | Q i | | M i |, If point ( , ), ,but |Q i |; d) The data set 16mm below the track vertex |Q i | Slicing process: Subtract the x-coordinates of two adjacent points; if the difference is greater than 0.3, then the point is used as the boundary of the set; ( , ), ( , ... ( , ), but |O 1 i |, |O 2 i | 2. The intelligent detection method for turnout spatial structure parameters according to claim 1, characterized in that, The optical path configuration of the laser scanning module is as follows: the blue laser camera is symmetrically tilted at 30° to scan the top surface of the rail; the red laser camera scans the rail waist and wheel flange groove area vertically downward.

3. The intelligent detection method for turnout spatial structure parameters according to claim 1, characterized in that, The sensor layout of the traveling frame is as follows: fiber optic gyroscopes and tilt sensors are installed at the geometric center of the longitudinal beam; displacement gauges are symmetrically arranged at both ends of the crossbeam to measure the dynamic changes in track gauge.

4. The intelligent detection method for turnout spatial structure parameters according to claim 1, characterized in that, The filtering circuit of the data acquisition module includes: track gauge signal: second-order Butterworth low-pass filter with a cutoff frequency of 10Hz; high and low signals: FIR low-pass filter RLS algorithm.

5. The intelligent detection method for turnout spatial structure parameters according to claim 1, characterized in that, The positioning structure is assembled as follows: the lateral positioning pin restricts the X-axis displacement, and the longitudinal positioning pin restricts the Y-axis displacement; the positioning disc achieves Z-axis clamping through conical surface mating.

6. A non-volatile storage medium, characterized in that, The non-volatile storage medium includes a stored program, wherein the program, when running, controls the device where the non-volatile storage medium is located to execute the method of claim 1.

7. A terminal device, characterized in that, The terminal device includes: a processor, a memory, a communication interface, and a bus; the processor, the memory, and the communication interface are connected through the bus and communicate with each other; the memory stores executable program code; the processor reads the executable program code stored in the memory to run a program corresponding to the executable program code, so as to execute the method as described in claim 1 above.

Images